SPITZER observations of Abell 1763. III. The infrared luminosity function in different supercluster environments

Abstract

We determine the galaxy infrared (IR) luminosity function (LF) as a function
of the environment in a supercluster at z=0.23, using optical, near-IR, and
mid- to far-IR photometry, as well as redshifts from optical spectroscopy. We
identify 467 supercluster members in a sample of 24-micron-selected galaxies,
on the basis of their spectroscopic (153) and photometric (314) redshifts. IR
luminosities, stellar masses and star formation rates (SFRs) are determined for
supercluster members via spectral energy distribution fitting and the Kennicutt
relation. Galaxies with active galactic nuclei are excluded from the sample. We
determine the IR LF of the whole supercluster as well as the IR LFs of three
different regions in the supercluster: the cluster core, a large-scale
filament, and the cluster outskirts (excluding the filament). The IR LF shows
an environmental dependence which is not simply related to the local galaxy
density. The filament, an intermediate-density region in the A1763
supercluster, contains the highest fraction of IR-emitting galaxies at all
levels of IR luminosities. As expected, the core contains the lowest fraction
of IR-emitting galaxies and almost no Luminous IR Galaxies (LIRGs). The
relation between galaxy specific SFRs and stellar masses does not depend on the
environment, and it indicates that most supercluster LIRGs (in particular those
in the filament) are rather massive galaxies with relatively low specific SFRs.
A comparison with previous IR LF determinations from the literature confirms
that the mass-normalized total SFR in clusters increases with redshift, but
more rapidly than previously suggested for redshifts <0.4. We interpret our
findings within a possible scenario for the evolution of galaxies in and around
clusters [Abridged].

Full text

Page 1

arXiv:1106.1657v3 [astro-ph.CO] 20 Jul 2011
Astronomy & Astrophysics manuscript no. lf.v8
July 21, 2011
c ? ESO 2011
SPITZER observations of Abell 1763. III. The infrared luminosity
function in different supercluster environments
A. Biviano1, D. Fadda2, F. Durret3,4, L.O.V. Edwards2, F. Marleau5
1INAF/Osservatorio Astronomico di Trieste, via G. B. Tiepolo 11, I-34131, Trieste, Italy
2NASA Herschel Science Center, Caltech 100-22, Pasadena, CA 91125, USA
3UPMC Universit´ e Paris 06, UMR 7095, Institut d’Astrophysique de Paris, 98bis Bd Arago, F-75014 Paris
4CNRS, UMR 7095, Institut d’Astrophysique de Paris, 98bis Bd Arago, F-75014 Paris, France
5Department of Astronomy and Astrophysics, University of Toronto, 50 Saint George Street, Toronto, ON M5S 3H4, Canada
Received / Accepted
ABSTRACT
Context. The study of galaxy luminosity functions (LFs) in different environments provides powerful constraints on the physics of
galaxy evolution. The infrared (IR) LF isa particularly useful tool since it is directly relatedto the distributionof galaxy star-formation
rates (SFRs).
Aims. We aim to determine the galaxy IR LF as a function of the environment in a supercluster at redshift 0.23 to shed light on the
processes driving galaxy evolution in and around clusters.
Methods. We base our analysis on multi-wavelength data, which include optical, near-IR, and mid- to far-IR photometry, as well as
redshifts from optical spectroscopy. We identify 467 supercluster members in a sample of 24-µm-selected galaxies, on the basis of
their spectroscopic (153) and photometric (314) redshifts. IR luminosities and stellar masses are determined for supercluster members
via spectral energy distribution fitting. Galaxies with active galactic nuclei are identified by a variety of methods and excluded from
the sample. SFRs are obtained for the 432 remaining galaxies from their IR luminosities via the Kennicutt relation.
Results. We determine the IR LF of the whole supercluster as well as the IR LFs of three different regions in the supercluster:
the cluster core, a large-scale filament, and the cluster outskirts (excluding the filament). A comparison of the IR LFs of the three
regions, normalized by the average number densities of r-band selected normal galaxies, shows that the filament (respectively, the
core) contains the highest (respectively, the lowest) fraction of IR-emitting galaxies at all levels of IR luminosities, and the highest
(respectively, thelowest) total SFRnormalized by optical galaxy richness. Luminous IRgalaxies (LIRGs)arealmost absent in thecore
region. The relation between galaxy specific SFRs and stellar masses does not depend on the environment, and it indicates that most
supercluster LIRGs are rather massive galaxies with relatively low specific SFRs. A comparison with previous IR LF determinations
from the literature confirms that the mass-normalized total SFR in clusters increases with redshift, but more rapidly than previously
suggested for redshifts<∼0.4.
Conclusions. The IR LF shows an environmental dependence that is not simply related to the local galaxy density. The filament, an
intermediate-density region in the A1763 supercluster, contains the highest fraction of IR-emitting galaxies. We interpret our findings
within a possible scenario for the evolution of galaxies in and around clusters.
Key words. Galaxies: luminosity function - Galaxies: clusters: general - Galaxies: clusters: Abell 1763 - Galaxies: evolution -
Galaxies: starburst
1. Introduction
The distribution of galaxy luminosities, i.e. the galaxy lumi-
nosity function (LF), and its environmental dependence have
often been used to provide strong constraints on theories of
galaxyevolution(see e.g.the investigationsofZucca et al. 2009;
Merluzzi et al. 2010; Peng et al. 2010). Galaxy environmentcan
beimportantinshapingseveralgalaxyproperties,suchas colors,
morphologies, and star-formation rates (SFRs; see e.g. Biviano
2008; Gavazzi 2009, for reviews). Since galaxySFRs are strictly
related to their total infrared (IR) emission (Kennicutt 1998),
powerful constraints on how galaxies evolve in relation to their
environmentare expectedto be obtained fromthe analysis of the
galaxy IR LFs.
Following early studies of the IR properties of cluster galax-
ies with IRAS and ISO (see Metcalfe et al. 2005, and references
therein), the Spitzer Space Telescope (Werner et al. 2004), the
Send offprint requests to: Andrea Biviano, biviano@oats.inaf.it
AKARI satellite (Murakami et al. 2007), and now the Herschel
Space Observatory (Pilbratt et al. 2010), have only recently al-
lowed a precise derivation of galaxy IR LFs at various redshifts
and in various environments in a precise way.
Most determinations of galaxy IR LFs in cluster and super-
cluster environments have so far been based on Spitzer data.
Bai et al. (2006, 2009) analyzed the IR LFs of the rich nearby
clustersComaandA3266.Accordingtotheiranalysis,thebright
end of the IR LF has a universal form for local rich clusters, and
cluster and field IR LFs have similar values of their character-
istic luminosities, L⋆
forming galaxy fraction than field galaxies, although this frac-
tion increases with cluster-centric distance.
IR. Nearby rich clusters have a lower star-
These results were confirmed by Finn et al. (2010), who
analyzed a larger sample of clusters in the redshift range
0.4 ≤ z ≤ 0.9. They noted the similarity in the shape of the clus-
terandfieldIR LFs,andconfirmedthatthereis anincreaseinthe
1

Page 2

Biviano et al.: A1763 IR luminosity function
fraction of luminous IR galaxies (LIRGs1) with cluster-centric
distance, out to 1.5 virial radii2, where it is still below the field
value. Temporin et al. (2009) also noted the absence of LIRGs
from the central regions of a sample of 32 X-ray selected clus-
ters. Recent Herschel observations provide evidence of a lack
of IR galaxies not only at the bright end but also the faint end
of the IR LF of the nearby Virgo cluster relative to the field
(Davies et al. 2010).
The environmental dependence of the fraction of high-SFR
galaxiesmaynotbe a simplefunctionofcluster-centricdistance.
Fadda et al. (2008, Paper 0) detected a large-scale filament3in
the IR, connecting a rich and a poor cluster at z ∼ 0.2. They
observed that the fraction of high-SFR galaxies is largest in
the filament, i.e. larger than in the cluster core, but also larger
than in other, lower density, regions of the supercluster. The fil-
ament detected by Fadda et al. (2008) was the first to be found
via IR observations (with Spitzer); Herschel observations have
recently revealed other large-scale structure filaments traced by
IR-emittinggalaxies(Haines et al.2010;Pereira et al.2010),but
the analysis of their galaxy populations is still ongoing.
Koyama et al. (2008, 2010) observed that the medium- and
low-densityregionsof another(moredistant,z ∼ 0.8)superclus-
ter host comparable fractions of star-forming galaxies, while red
mid-IRemittersarepreferentiallylocatedinmedium-densityen-
vironments, such as galaxy filaments. Both Fadda et al. (2008)
and Koyama et al. (2010) argued that star-formation is triggered
in galaxies in the infall regions around clusters. Gallazzi et al.
(2009) came to the same conclusion after analyzing the IR
galaxy population in a z = 0.165 supercluster. They also found
that while the IR galaxies prefer to live in medium-density en-
vironments, their SFRs are not particularly high for their stel-
lar masses (M⋆), i.e. they have normal specific SFRs (sSFR ≡
SFR/M⋆).
Groups are anotherenvironmentcharacterized,as in the case
of filaments, by galaxy densities intermediate between cluster
cores and the field. Tran et al. (2009) determined the IR LFs of
a rich galaxy cluster and four galaxy groups at z ∼ 0.35. The
fraction of galaxies with a high SFR was found to be four times
larger in the groups than in the cluster, or equivalently,the group
IR LF has an excess at the bright end relative to the cluster IR
LF. On the basis of this result, Chung et al. (2010) interpreted
the excess of bright IR sources in the IR LF of the Bullet cluster
(z ∼ 0.3) as being due to the galaxy population in an infalling
group (the “bullet” itself).
The IR LF not only depends on the environment, but also
on redshift. Bai et al. (2009) compared the average IR LFs of
two nearby (z ≤ 0.06) and two distant (z ∼ 0.8) clusters
(using the data of Bai et al. 2006, 2007). They concluded that
there is an evolution with z of both the characteristic luminos-
ity L⋆
z clusters contain more and brighter IR galaxies. This evo-
lution of the cluster IR LF results in a rapid increase with
IRand the normalization of the LF, n⋆, such that higher-
1LIRGs are galaxies with a total (8–1000 µm) IR luminosity
LIR≥ 1011L⊙.
2The cluster virial radius, r200, is the radius within which the en-
closed average mass density of a cluster is200 times the critical density.
The circular velocity v200is defined as v200 = 10H(z)r200. The virial
mass M200follows from the two previous quantities, M200= r200v2
3In Paper 0, we originally identified two filaments, running almost
parallel in projection in the sky, but slightly separated along line-of-
sight velocity space. Subsequent spectroscopic observations indicate
that the two filaments merge into one at large distances from the A1763
cluster core. For simplicity, we therefore here refer to a single filament
in the supercluster.
200/G
z in the total SFR of cluster galaxies divided by the total
cluster mass, ΣSFR/mass ∝ (1 + z)5.3, a result anticipated by
Geach et al. (2006), who suggested an even faster evolution.
Another way to characterize this evolution is to look at the
fraction of IR-emitting galaxies (above a given IR luminos-
ity, LIR) as a function of z. This fraction is observed to in-
crease with z, a phenomenon called “the IR Butcher-Oemler ef-
fect” (Saintonge et al. 2008; Haines et al. 2009a; Temporin et al.
2009), since it is reminiscent of the increasing fraction of blue
cluster galaxies with z (Butcher & Oemler 1984). The increas-
ing fraction of LIRGs with z appears however to be a common
phenomenonin clusterandfieldenvironments(Finn et al.2010).
To shed light on the physical processes responsible for the
environmentaland redshift dependence of the IR LF, we present
a study of the IR LF of galaxies in the z = 0.23 A1763–A1770
supercluster. Our analysis is restricted to the part of the super-
cluster that includes the rich cluster A1763, part of the filament
connecting the two clusters (see Paper 0), and the outskirts re-
gion aroundthe A1763 cluster core, excludingthe filament itself
(see Sect. 3.3).
In Sect. 2, we describe our observationaldata-set (Sect. 2.1),
assign supercluster memberships to the observed IR galaxies
(Sect. 2.2), and determine their total IR luminosities (Sect. 2.3)
and stellar masses (Sect. 2.4). In Sect. 3, we describe the correc-
tions applied to the IR galaxy counts (Sect. 3.1) to determine the
supercluster IR LF (Sect. 3.2). We then determine the corrected
IR LFs ofthreedifferentregionsoftheA1763superclusterto ex-
plore environmental effects (Sect. 3.3). We compare our results
with previous results from the literature in Sect. 3.4. In Sect. 4,
we discuss our results and summarize them in Sect. 5.
We adopt H0 = 70 km s−1Mpc−1, Ωm = 0.3, ΩΛ = 0.7
throughout this paper. In this cosmology, 1 arcmin corresponds
to 222 kpc at the cluster redshift.
2. The data set
2.1. Observations
The data used in this study were obtained as part of a multi-
wavelength observational campaign conducted with several
space- and ground-based telescopes. Details are provided in
Edwards et al. (2010a, Paper 1). Here we summarize the main
characteristics of the data set. A field of ∼ 40 × 55 arcmin2cen-
tered on the A1763 cluster was covered by MIPS 24, 70, and
160 µm observations from Spitzer. Two similar fields were also
covered by IRAC 3.6, 4.5, 5.8, and 8.0 µm observations from
Spitzer. A similar area was observed with the Palomar 200 inch
telescope in the r′, J,H, and Ksfilters. In addition, we obtained
spectroscopic observations for galaxies across the supercluster
region, using the KPNO WIYN and TNG telescopes (paper in
preparation). Finally, the A1763 field was covered by the Sloan
Digital Sky Survey (SDSS hereafter) in the u′,g′,r′,i′,z′photo-
metric bands, and we collected all data available in the A1763
field from the SDSS Seventh Data Release (DR7 hereafter). We
use Petrosian magnitudes and total fluxes in the following anal-
yses.
Our sample contains 10876 objects identified at 24 µm in
the MIPS field. The observational technique as well as the depth
of our MIPS observations are very similar to those of the “verifi-
cation survey” in the Spitzer Space Telescope Extragalactic First
Look Survey (EFLS hereafter; Fadda et al. 2006). For this rea-
son, we assume that the completenessandpurityfunctionsof the
EFLS and those of our survey are identical. This is a conserva-
tive assumption because the EFLS sources were selected using
2

Page 3

Biviano et al.: A1763 IR luminosity function
Fig.1. The rest-frame velocities versus cluster-centric distances
of the galaxies with available z. The vertical solid pink line
marks the distance to the cluster virial radius, r200. Black dots
represent the 357 galaxies selected as supercluster members by
the algorithm of Mamon et al. (2010), interlopers are marked by
X’s. Red circles identify 24 µm emitters. 153 of them are se-
lected as supercluster members.
the peak signal-to-noise ratio, while here sources were selected
using the aperture signal-to-noise ratio, which is more efficient
in the rejection of false detections. Completeness, Cdet, is de-
fined as the fraction of real sources that are detected, and purity,
Pdet, is defined as the fraction of real sources among the detected
ones. The completeness is Cdet∼ 80% at 24 µm flux densities
f24= 0.2 mJy, and close to 100% at f24> 0.4 mJy. The purity is
Pdet∼ 95%atf24≥ 0.2mJyandabove(seeFig.13inFadda et al.
2006).
We base the determination of the IR LFs on the sample of
24 µm-detected IR-emitting galaxies, since our 70 µm and 160
µm observations are not as deep. About 60% of the 24 µm-
selected objects have f24≥ 0.2 mJy and therefore belong to the
sample with ≥ 80% completeness and ∼ 95% purity. We use
these completeness and purity estimates in the construction of
the supercluster IR LF (see Sect. 3.1).
2.2. Supercluster membership
To define the supercluster membership of the galaxies in the
cluster field, we use both spectroscopic (z) and photometric red-
shifts (zp).
We use the algorithm of Mamon et al. (2010) to identify the
supercluster members among the galaxies with available z. This
algorithm tries to infer the galaxy cluster membership from the
location of the galaxy in the cluster-centric distance – velocity
diagram shown in Fig. 1, based on the modeling of the mass
and anisotropy profiles of cluster-sized halos extracted from a
cosmological numerical simulation. The procedure is more ef-
fective than traditional approaches (e.g. Yahil & Vidal 1977) in
rejecting interlopers, while still preserving cluster members.
The galaxy rest-frame velocities with respect to the
cluster mean velocity are obtained from the usual rela-
tion v = c(z − z)/(1 + z) (Harrison & Noonan 1979), where
z = 0.2314 is obtained using the biweight estimator (Beers et al.
1990). We then obtain the galaxy projected distances from
Fig.2. SDSS DR7 photometric redshift zp (Artificial Neural
Network estimates, Oyaizu et al. 2008) versus spectroscopic
redshift z for the IR-emitting galaxies with available z and zp
in the supercluster field (471 galaxies in the displayed z and zp
ranges). The solid pink line is the identity relation z = zp. The
dash-dotted red lines indicate the chosen zprange for member-
ship selection (see text and Fig. 3).
the cluster center, defined by its X-ray peak emission,
RA=13h35m17.96s, δ=40◦59′55.8′′(Cavagnolo et al. 2009).
The algorithm of Mamon et al. (2010) requires initial es-
timates of the virial radius, r200, and circular velocity, v200,
which we obtain from the cluster velocity dispersion estimate of
Paper 0, by following Mauduit & Mamon (2007, Appendix A),
and using the relation of Gao et al. (2008) to infer the concentra-
tion of the cluster mass-density distribution.
We run the procedure on the whole sample of 1364 ob-
jects with available redshift estimates in the supercluster field.
The procedure is run iteratively until convergence on the num-
ber of selected members. We identify 357 supercluster mem-
bers (they are shown as filled dots in Fig. 1). Other algorithms
(e.g. den Hartog & Katgert 1996; Fadda et al. 1996) lead to very
similar membership definitions. The average cluster redshift and
velocity dispersion determined for this sample of supercluster
members are z = 0.2315 ± 0.0003 and σv = 1051+51
We use these values to estimate the cluster virial radius and cir-
cular velocity as before, finding r200 = 2.066 Mpc and v200 =
1623 km s−1, which do not differ significantly from the initially
adopted values.
Of the 357 identified supercluster members, 153 are 24 µm-
emitters.
To estimate the supercluster membership for the subset of
galaxies without z, we rely on zp-estimates. We consider six dif-
ferent zp-estimates for the galaxies in our sample. In particu-
lar, we consider the ANNz (Collister & Lahav 2004) and EAZY
(Brammer et al. 2008) algorithms, as well as a χ2minimiza-
tion fitting of the spectral energy distribution (SED, hereafter)
of the galaxies in our sample using SED model templates from
−54km s−1.
3

Page 4

Biviano et al.: A1763 IR luminosity function
Fig.3. The zp distributions for the IR-emitting galaxies with
available z in the supercluster field. The solid black (respec-
tively, dashed blue) histogram represents the zpdistribution for
the galaxies selected as members (respectively, not selected as
members) on the basis of their z. The two vertical red dash-
dotted lines identify the lower and upper zplimits used to iden-
tify supercluster members in the sample of galaxies without z
(not shown here).
Polletta et al. (2007). We also consider the three zpestimates di-
rectly available from the SDSS DR7. Of these six zpestimators,
we finally adopt one of those provided in the SDSS DR7, that
based on the Artificial Neural Network technique (Oyaizu et al.
2008). This estimator providesthe tightest correlationbetween z
and zpfor the subsample of galaxies in the A1763 field that have
both quantities available (see Fig. 2).
To select the supercluster members in the sample of 24 µm-
emitters on the basis of their zp, we define a zp-range around
the mean supercluster redshift. The lower and upper zp-limits
that define this selection range must be chosen in such a way as
to maximize the number of real supercluster members with zp
within these limits, and, at the same time, minimize the number
of background and foreground galaxies that also happen to have
their zpwithin these limits. The choice of these zp-limits can
only be based on the sample of galaxies with zpand z, so that
we can perform the most robust zp-based membership selection
possible based on the well-established spectroscopic member-
ship.
We proceed as follows. We assume that the 153 supercluster
membersselectedonthebasis oftheirz areall realmembers.We
then determine the zp-distribution of these 153 galaxies (shown
asa solidblackhistograminFig.3),as wellas thezp-distribution
of the galaxies with z in either the foregroundor the background
of the supercluster (dashed blue histogram in Fig. 3). Using the
whole sample of galaxies with z and zp, we define the purity
and completeness to be, respectively4: Ppm ≡ Npm∩zm/Npm∩z
and Cpm≡ Npm∩zm/Nzm∩p, where Nzm∩pis the number of spec-
troscopically confirmed cluster members with available zp, and
Npm∩z(respectively, Npm∩zm) is the number of galaxies with z
(respectively, the number of spectroscopically confirmed clus-
ter members) that have zpwithin a given zp-range. Following
Knobel et al. (2009) we determine the optimal zprange by mini-
4For the sake of simplicity hereafter, we use the letter “p” in lieu of
“zp” in the subscripts.
Fig.4. Completeness (Cpm) and purity (Ppm) of the sample of
IR-emitting supercluster members selected on the basis of their
zp(0.166 ≤ zp ≤ 0.290), as a function of f24. Cpmand Ppmare
estimated using the sample of 24 µm-emitters with both z and
zpavailable, and assuming the members selected on the basis
of their z are all real members. The black X’s are for the total
sample. The red dots, blue squares, green stars are for the core,
filament, and outskirts subsamples, respectively (see 3.3 for the
definition of these subsamples). Horizontal bars indicate the f24
bin intervals. Vertical bars indicate 1-σ uncertainties.
mizing
Cpm = 0.73 and Ppm = 0.42, corresponding to the zp-range
0.166–0.290. The dependence of Cpmand Ppmon f24is not very
strong (see Fig. 4). Among the galaxies without z, 314 have zp
within this range.
In Fig. 2, the two red dashed lines indicate the chosen zp-
range. It can be seen that most of the supercluster galaxies fall
in that range, but also many of the galaxies that belong to two
other z-peaks, one at z ∼ 0.17, another at z ∼ 0.29. We consider
whether it is possible to increase the purity of the sample of zp-
selected cluster members by identifying and then removing the
galaxy structures responsible for these two z-peaks. The lower-z
peak does not correspond to a concentrated structure in space.
The higher-z peak does seem to correspond, at least partly, to
a spatial concentration of galaxies, located at the edge of the
observed Spitzer field. However, removing the (small) region
correspondingto this (presumed) galaxy concentration from our
analysis has hardly any noticeable effect on the results presented
in this paper.
In total, we select 467 IR-emitting galaxies as supercluster
members, 153 on the basis of z, 314 on the basis of zp. We base
the derivation of the supercluster IR LF on both the total sam-
ple of members (the z ∪ zpsample, hereafter), and the sample
of z-selected members (the z sample, hereafter; see Sect. 3.2).
Using both samples allows us to check the influence of possible
systematic errors because the z ∪ zpsample is affected by sig-
?
(1 − Ppm)2+ (1 − Cpm)2. The minimum is obtained for
4

Page 5

Biviano et al.: A1763 IR luminosity function
Fig.5. The cumulative fractions of galaxies of different SED
classes as a function of LIR, for the z ∪ zpsample. The pink,
blue,green,andblack-shadedregionscorrespondtothefractions
of SBGs, PSBGs, SFGs, and ETGs, respectively (as labeled).
The orange and red-shaded regions correspond to the fractions
of SED-identified AGNs (mostly at low LIR) and AGNs identi-
fied in Paper 2 from X-ray or radio emission, respectively.
nificant contamination by non-real members (low purity), while
the z sample is affected by larger incompleteness than the z ∪ zp
sample.
2.3. Total infrared luminosities
To determine the total IR luminosities (LIR) of the 467 su-
percluster members, we fit the galaxy SEDs with two sets of
modeltemplates,onefromGRASIL(Silva et al.1998),theother
from Polletta et al. (2007). These templates span a wide range
of galaxy types, with different formation redshifts, and were
used in Paper 0 as well as (in part) in Biviano et al. (2004) and
Coia et al. (2005a,b). In total, we consider 61 SED templates of
galaxies of different ages and types, belonging to the following
five classes:
– ETGs, early-type galaxies;
– SFGs, normal star-forming galaxies;
– SBGs, starburst galaxies;
– PSBGs, post-starburst galaxies;
– AGNs, active galactic nuclei.
We find the best-fit templates by comparingthe template and
observed fluxes via a χ2minimization procedure. To compute
the template fluxes in the observed photometric bands, the tem-
platesareredshiftedtothegalaxy(photometricorspectroscopic)
redshifts andconvolvedwith the filter responsecurves.The min-
imization procedureis run interactively, allowing, when needed,
the eye-rejection of deviant photometric data in the fits of indi-
vidualgalaxySEDs. We finally determineLIR by integratingthe
best-fit model SEDs over the 8–1000µm rest-frame wavelength
range.
Given the galaxy IR luminosities, we determine the
galaxy SFRsusing the relation
SFR[M⊙/yr] = 1.7 · 10−10· LIR/L⊙. This relation is clearly
valid only when a galaxy IR luminosity is not dominated by the
emission from an AGN. Since most galaxies in our sample lack
of Kennicutt(1998),
Fig.6.
LIR,SED/LIR,R09 versus LIR,SED. Different symbols identify
different galaxy SED classes. Black crosses for early-type
galaxies (ETGs), red X’s for active-galactic nuclei (AGNs),
pink diamonds for starburst galaxies (SBGs), blue squares for
post-starburst galaxies (PSBGs), and green circles for normal
star-forming galaxies (SFGs). The dashed line is the biweight
average ratio of the sample of non-AGN galaxies.
Comparisonoftwo IR luminosityestimates,
far-IR photometry,it may be difficult for us to distinguish AGNs
from galaxies with IR emission dominated by star formation. It
is therefore also worth considering other AGN diagnostics.
In Edwards et al. (2010b, Paper 2), we identified AGNs in
the A1763 region using optical, radio, X-ray data, and IRAC
colors. Nine of the AGNs identified in Paper 2 are in oursample,
and only one of them has been classified as an AGN based on its
SED. This is unsurprising, since AGNs become visible in dif-
ferent bands at different stages of their evolution (Hickox et al.
2009), and since the AGNs identified in Paper 2 in the IRAC
colordiagramare at the marginof the AGN-identificationregion
(see Fig. 6 in Paper 2). We also adopt the AGN classification
of Paper 2 for the 8 galaxies with non-AGN SED classification,
bringing the total of AGNs in our sample to 35 (13 with avail-
able z). We are therefore confident we have identified most (if
not all) galaxies with AGN-dominated IR emission.
The relative contribution of the different SED classes in dif-
ferent LIR bins is shown in Fig. 5 for the z ∪ zpsample (the
equivalent figure for the z sample is very similar and not shown
here). Fig. 5 shows that SBGs contribute mostly at high LIR, but
a significant fraction of the LIRGs are normal SFGs. The frac-
tion of SFGs and of PSBGs increases at lower LIR, and SFGs
dominate at intermediate LIR. Most of the galaxies at the faint-
end of the IR LF are ETGs. In line with previous results and
with our previous analysis (Paper 2), we find the contribution
of AGNs to the IR LF of A1763 to be small (e.g. Geach et al.
2009; Krick et al. 2009; Chung et al. 2010), and to increase with
LIR(e.g. Bothwell et al. 2011; Goto et al. 2011).
In order to check the robustness of our SED-based LIR es-
timates we consider alternative estimates based on direct rela-
tions betweenf24andLIR, fromRieke et al. (2009) and Lee et al.
(2010). When comparing the different LIR estimates, we only
consider the subsample of 140 spectroscopicallyconfirmednon-
AGN A1763 members, to be sure that the comparisons are un-
affected by the additional scatter introduced by photometric red-
5